Fix bug in Date.to_timedelta

[xylar]

This merge fixes a bug in Date.to_timedelta where the local variable day is used instead of self.day.

Also, fix some PEP8 issues in Date.

[xylar]

No existing code (except our CI tests) uses Date.to_timedelta so analysis should be unaffected by this bug.

My tests on an ACME G run with MPAs at QU240km resolution show that the results are bit-for-bit unchanged. (I did not test sea ice climatology plots because I don't have the mapping files for this particular grid.)

[xylar]

@milenaveneziani, this should hopefully be an easy review. I may update the CI test because it somehow didn't catch this bug so probably could use a more rigorous test.

[xylar]

I added the more rigorous test.

[pwolfram]

Assumes that there are no leap years, which is not general and probably needs fixed in #82 for the case where leap years may be allowed for MPAS.

[pwolfram]

@xylar, regarding

No existing code (except our CI tests) uses Date.to_timedelta so analysis should be unaffected by this bug.

this doesn't seem to be the case, e.g., https://github.com/MPAS-Dev/MPAS-Analysis/blob/develop/mpas_analysis/shared/mpas_xarray/mpas_xarray.py#L189. This could have unexpected side effects for the case of 'timedelta64[ns]'-based dataset time dimensions, which are not being tested via the QU240km. However, in all fairness this is clearly not the 85% of cases we routinely consider.

[xylar]

@pwolfram, I think you're maybe confusing Date.to_timedelta with pandas.to_timedelta, which are not related functions (other than that both produce a timedelta object). mpas_xarray is fully unaware of the existence of the Date class, so this change can't affect it. Grepping through the code in the current develop, I see:

$ grep -r "to_timedelta" .
./mpas_analysis/shared/timekeeping/Date.py:            raise ValueError("self.isInterval == True. Use to_timedelta "
./mpas_analysis/shared/timekeeping/Date.py:    def to_timedelta(self):
./mpas_analysis/shared/timekeeping/Date.py:                             "instead of to_timedelta")
./mpas_analysis/shared/mpas_xarray/mpas_xarray.py:                     pd.to_timedelta(time_var.values, unit='ns')]
./mpas_analysis/test/test_date.py:        timedelta1 = date.to_timedelta()

The only call to Date.to_timedelta is in the testing infrastructure.

[pwolfram]

@xylar, thanks for clarifying.

[xylar]

Thanks @pwolfram!